Predibase

core-platform

closed(proprietary post-training stack + turbo serving)

source

partial(LoRAX and Ludwig are adjacent OSS

OSS-components

LoRAX(open source multi-LoRA serving)+Ludwig(open source)

runs-on

Predibase-SaaS-or-customer-VPC

license

Proprietary(managed) with Apache-licensed LoRAX/Ludwig adjuncts

The founders created and maintain OSS LoRAX (multi-LoRA serving) and Ludwig, but the Predibase fine-tuning/RFT platform, which is the thing being scored, is proprietary managed SaaS running on Predibase's own cloud or a customer VPC. The OSS sits at the serving and declarative-ML edges. Class corrected from open_core on 2026-07-30. In this ladder open_core means an OSI core with functionality withheld for a paid tier, and there is no open core here, only an open periphery. A repo that exists while the runtime does not ship is source:partial, which the ladder scores 2/source_available. The score itself was already 2.

• https://docs.predibase.com/ recorded 2026-06-04

  managed platform to fine-tune & serve open source LLMs; SFT/RFT/continued-pretraining; built by LoRAX creators; SaaS/VPC

• https://www.rubrik.com/company/newsroom/press-releases/25/rubrik-to-acquire-predibase-to-accelerate-agentic-ai-adoption recorded 2026-06-04

  Rubrik acquisition (Jun 25 2025); proprietary post-training stack + turbo serving + OSS LoRA eXchange

No hard per-platform usage figure (jobs / customers) disclosed this run. Reported traction: enterprise post-training/agent platform now distributed under Rubrik (acquisition valued $100M-$500M per press), founder-pedigree (Google/Uber), and a widely-adopted OSS serving layer (LoRAX). Placed at 2 on reported enterprise traction, not a verified usage_volume number; low confidence pending hard data.

• https://techcrunch.com/2025/06/25/rubrik-acquires-predibase-to-accelerate-adoption-of-ai-agents/ recorded 2026-06-04

  Rubrik acquires Predibase (Jun 25 2025) to scale enterprise AI-agent/fine-tuning adoption

• https://docs.predibase.com/ recorded 2026-06-04

  production private serverless endpoints; SaaS/VPC enterprise deployment

Strong managed method depth - notably the pioneer of end-to-end RFT for LLMs - plus efficient multi-LoRA serving over open models. Scored 4: matches the RFT-capable managed services (OpenAI/Azure/Fireworks at C4) and adds best-in-class multi-adapter serving, but is still a black-box managed platform without exposed large-scale training internals, so below the OSS scale-frontier definers (Megatron-LM C5).

• https://www.rubrik.com/blog/ai/25/introducing-reinforcement-fine-tuning-on-predibase recorded 2026-06-04

  first end-to-end reinforcement fine-tuning platform for LLMs (per Rubrik; corroborated by acquisition coverage)

• https://docs.predibase.com/ recorded 2026-06-04

  SFT, RFT, continued pretraining; LoRAX multi-adapter serving; turbo serving 2x